Many scenes included several characters within a single frame, which is still challenging for AI models. In such cases, I refined strong outputs manually using Photoshop — adjusting clothing, adding or removing objects, correcting facial expressions, and fixing other visual defects.